Readers can help me choose the best metaphor. (1) A snowball which starts the size of a pea but gains in strength and speed as it rolls downhill, mindlessly consuming all in its path; or (2) An avalanche, a furious, powerful deluge which is set off by some small thing, mindlessly consuming all in its path.
The application is to fMRI and other imaging studies of the brain which show that some “aberrant” (i.e. non-leftist) behavior can be pinned to some small region inside the skull. These statistical studies, which were nonexistent just a short while ago, now pop up weekly. Each works cites the others, and in so doing hopes to convince by sheer mass. Like a snowball or avalanche. Call it Death by Correlation, or Attack of the P-Value.
We have seen dark triads, enjoyment of custom, faith, and many, many others, all said to be caused by quirks in the brains of conservatives and theists. Today’s example is “Being ‘Born-Again’ Linked to More Brain Atrophy.” Well, at least this is clear enough. Rotting, shriveling brains “linked” to faith.
The abstract of the peer-reviewed paper by Amy Owen et alia is instructive. It opens,
Despite a growing interest in the ways spiritual beliefs and practices are reflected in brain activity, there have been relatively few studies using neuroimaging data to assess potential relationships between religious factors and structural neuroanatomy.
That sentence is paradigmatic: “Despite interest in X, we have not yet seen work in subgroup Y” opens many papers. It suggests a bandwagon groaning under the weight of scientists with hyperactive pituitaries, or whatever cranial organ is responsible for copycat research—and according to these folks some such organ must exist.
As depressing as this beginning is, let’s push on. Our data-mining discovery-of-the-day comes from comparing “religious factors and hippocampal volume change using high-resolution MRI data of a sample of 268 older adults.” Main claim: “Significantly greater hippocampal atrophy was observed for participants reporting a life-changing religious experience.” And there you have it: withered brains produce faith. In case the connotation wasn’t obvious, the press article reminds, “Shrinkage of the hippocampus is also associated with depression, dementia and Alzheimer’s disease.”
(You know, it’s not often I brag or boast or plead, but I should receive some kind of humanitarian award for reading these papers so you don’t have to.)
Participants, with double the number of women, mean age near 70, were measured twice, a baseline and follow-up—follow-ups were not consistent: times “ranged from 2—8 years.” Participants were “those meeting DSM-IV criteria for major depressive disorder” and those “never-depressed”. They never tell us how many in this study were depressed and how many not. Ah, who needs rigor? This is statistics!
Affiliation was queried: non-born-again, born-again Protestant, Catholic, and Other. Only 19—just 19; a mere 19; 7%—listed no religion. Those claiming not to be born again were asked whether they had had “life-changing religious experiences” (LCRE).
Now, over the course of follow-up, twenty-two—that is, 22, which is 1 more than 21—people were newly born again and 23 others had new LCREs. That is 55 total, folks, which is nearly three times as many who claimed no religion prior to the study. That must mean that many people merely reported switching their religious label. How did the researchers handle this ambiguity in their statistical models?
(I really am going to have to devise a graphic which indicates crickets chirping.)
Understand, dear reader: the evidence for this study is entirely statistical, the result of regression models producing publishable p-values. They created two models for change in (falsely assumed error-free) measures of the left and right hippocampus. For the LH model, every measure of religion, pre or post, gave negative model coefficients except “Other” (but that had an unacceptable p-value). This includes “None,” which I remind us means no religion. Same for the RH model, except “Other” was negative here; “None” was still negative. Negative means the item shrank the hippocampus.
Unfortunately for consistency in reporting, newly born-again status was not significant in either model; neither were new LCREs (baseline for both was). And do you know what? The authors did not report any kind of baseline comparison of hippocampus size and religious belief. Must have slipped their minds—which cranial organ was responsible for this error we do not yet know; further studies might tell us.
Now, participants were old. Yet, somehow, the model coefficient for “Duration in study” was positive for the LH model and negative for the RH model. How can this be? Aging causes the left but not the right hippocampus to grow? By golly, Owen has discovered that enrollment in fMRI studies encourages hippocampus growth! On the left side only, alas; the p-values aren’t there yet, but this is early days: more research is needed. Send donations.
As always, the most fun is had by reading the discussion, where authors allow their minds to range freely over possible explanations of the data. I was going to summarize these curious cogitations, but I quailed after reading, “Research on temporal lobe epilepsy indicates that features of hyper-religiosity may be positively associated with hippocampal atrophy, but findings are mixed.” Good grief. I had not the strength to continue.
But I did check: they hadn’t even an inkling that their results were a statistical artifact or due to a sloppy experimental protocol and even sloppier analysis. I blame myself and other statisticians for this. I really do.